testing the validity of the linder hypothesis for croatia...leontief (1953) found paradox in the...

62

Croatian Review of Economic, Business and Social Statistics (CREBSS)

UDK: 33;519,2; DOI: 10.1515/crebss; ISSN 1849-8531 (Print); ISSN 2459-5616 (Online)

Vol. 4, No. 1, 2018, pp. 62-73

Testing the validity of the Linder hypothesis

for Croatia

Hrvoje Jošić

University of Zagreb Faculty of Economics and Business, Croatia

[email protected]

Matej Metelko

University of Zagreb Faculty of Economics and Business, Croatia

[email protected]

Abstract This paper presents empirical evidence on the validity of the Linder hypothesis in the

case of Croatia. According to the Linder hypothesis, one of the new theories of

international trade, countries with a similar level of income per capita should trade

more. In order to investigate the trade pattern of Croatia's international trade, a

panel regression model is formulated including 184 Croatia's import partner countries

in the period from 2000 to 2016. The Linder effect was displayed and calculated

using the Linder variable expressed as an absolute difference between GDP per

capita of the importing and the exporting country. The cross-country panel

regression model is estimated using Pooled OLS, Fixed and Random effects models.

Results of the analysis have shown that the validity of the Linder hypothesis for

Croatia cannot be accepted. Instead, the structure of Croatia's trade is in line with

the gravity model of international trade.

Keywords: Croatia, gravity model of international trade, Linder hypothesis, panel

data.

JEL classification: C33, F11.

DOI: 10.2478/crebss-2018-0006

Received: May 12, 2018

Accepted: June 12, 2018

Introduction Early economists asked fundamental questions why countries trade, what are the

gains from trade and what determines patterns of international trade. Mercantilists

believed that international trade is zero sum game where only exporting country has

gains from trade at the expense of importing country. In that time international trade

has not yet been recognized as mutually beneficial for all countries involved in the

process of trade. Smith (1776) formulated a theory of absolute advantages. He

recognized international trade as mutually beneficial for both countries involved in

trade but could not explain trade pattern when one country has absolute

advantages in production of all goods. Ricardo (1817) created the theory of

comparative advantages which stated that country will specialize in production of

good in which it have greater relative productivity of labour. Shortcoming of

63



Vol. 4, No. 1, 2018, pp. 62-73

classical theories of international trade was the restrictive assumption of labour as

the only factor of production and explanation of patterns of international trade only

from the supply side. Heckscher-Ohlin theory, Heckscher (1919) and Ohlin (1933),

argue that relatively capital-abundant country will have comparative advantage

and specialize in production and export of capital-intensive good while relatively

labour-abundant country will have comparative advantage and specialize in

production and export of labour-intensive good. In that framework international

trade would happen between rich and poor country. Heckscher-Ohlin theory is

arguably the most empirically tested theory of international trade, but empirical

results do not give firmly support for this theory. According to Trefler (1995) the

precision of this theory in explaining patterns of international trade is no greater than

a coin toss.

Leontief (1953) found paradox in the Heckscher-Ohlin theory while attempting to

test it empirically on the data for United States. He found that United States

specialized in exports of labour-intensive goods and import capital-intensive goods,

which is contrary to the Heckscher-Ohlin theory. On the other side, Linder (1961)

proposed a possible resolution of the Leontief paradox that questioned Heckscher-

Ohlin theory. He stressed the importance of demand side arguing that trade will

happen between countries of similar demand structures.

Goal of this paper is testing the validity of Linder hypothesis for Croatia using Linder

variable representing absolute difference between trading partner countries gross

domestic product per capita and gravity model of international trade. Hypothesis of

the paper is:

H1... “Croatia’s imports trade pattern complies to Linder hypothesis contrary to

gravity model of international trade”.

Hypothesis of the paper will be tested using cross-country panel regression analysis

on the data for Croatia and its import partners from 2000 to 2016. Paper is structured

and organized in six chapters. After the introduction second chapter elaborates

theoretical aspects of Linder hypothesis. In the third chapter literature review is

presented while methodology and data are explained in the fourth chapter. Fifth

chapter displays empirical analysis, results and discussion. Final chapter exhibits

concluding remarks.

Theoretical aspects of Linder hypothesis Linder theory is demand-side oriented which is in the contrast with the supply-side

oriented classical theories of international trade. Linder challenged Heckscher-Ohlin

theory explaining that it ignored demand-related factors which are important in

explaining patterns of international trade. His prediction is that the most of trade

should occur between countries of similar demand structures and similar level of

economic development. Linder coined his famous hypothesis (Linder, 1961) which

stated that “the more similar the demand structure of the two countries the more

intensive potentially is the trade between these two countries“. Linder theory is one

of several „new“ theories, which have arisen after Leontief testing of Heckscher-Ohlin

theory. Some of the rigid assumptions of the Heckscher-Ohlin theory have been

questioned out. Other new theories of international trade include gravity model of

international trade, Kravis availability theory, theory of intra-industry trade, product

life-cycle theory and theory of competitive advantage. Linder variable representing

Linder effect catches differences in per capita income between countries implying

higher expected trade with smaller difference in national incomes. Per capita

income is used as proxy for preferences. According to Linder domestic exporters will

export to countries with similar trade preferences as domestic country (Kennedy,

64



Vol. 4, No. 1, 2018, pp. 62-73

McHugh, 1983). Heckscher-Ohlin theory assumed production and trade of capital-

intensive and labour-intensive products while Linder theory made a difference

between primary and industrial (manufactured) goods. Pattern of trade in

Heckscher-Ohlin theory considered trade between developed and developing

countries (North-South trade) while Linder predicted that the most of trade will occur

between developed countries (North-North trade). Critics of Linder theory argued

that although Linder explained pattern of trade in manufactured products, he did

not explain that trade can go in both ways (intra-industry trade). Furthermore, his

theory was designed only for developed economies, but not for developing

economies.

Research overview Kennedy and McHugh (1980) examined Linder hypothesis for 14 countries in the

years 1960 and 1975 using Pearson and Spearman correlation coefficients. The test

was a one-tailed test of non-negative correlation versus negative correlation from

which it was clear that there was absolutely no evidence of a Linder type effect in

explaining trade patterns. The majority of the coefficients had the wrong signs.

Kennedy and McHugh (1983) investigated Linder hypothesis for United States in the

years 1963, 1970 and 1976 using one-digit SITC manufacturing data. The results gave

no support to existence of Linder hypothesis and disapproved Heckscher-Ohlin

theory of trade so authors instead of two competing theories suggested

implementation of two rival theories into one general theory which would better

explain trade patterns. In this way new theory would combine „supply“ and

„demand“ sides of trade theories to better explain trade patterns. Arnon and

Weinblatt (1998) tested Linder theory for trade between developed and developing

countries. Contrary to economic theory of 1960's they prove validity of Linder

hypothesis for both developed and developing countries. McPherson, Redfearn and

Tieslau (2001) assayed the validity of Linder hypothesis for six East African countries.

For five of them they found an evidence in support of the hypothesis except in the

case of Tanzania. The analysis was important because Linder hypothesis was tested

exclusively on developing countries and not developed ones.

Choi (2002) tested Linder hypothesis for 63 countries in the period from 1970 to

1992. The results of the analysis pointed to the conclusion of accepting Linder

hypothesis. Values of Linder variable showed tendency in growth since 1990’s which

can be attributed to process of globalization and trade liberalization. Bukhari et al.

(2005) investigate the validity of Linder hypothesis for three South Asian countries,

namely Bangladesh, India, and Pakistan. The empirical results of the fixed-effects

panel data model pointed out to the conclusion that aforementioned developing

countries trade intensively with countries of similar development level. Although this

research was not applicated to all developing countries, it gives certain setting for

testing the validity of Linder theory for entire developing world.

In table 1 is presented literature review of empirical testings on the validity of

Linder hypothesis.

65



Vol. 4, No. 1, 2018, pp. 62-73

Table 1 Literature review of empirical testings on Linder hypothesis Author (Year) Data Sample Methodology Research results

Kennedy and

McHugh (1980) 1960 and

1975 14 countries Pearson and

Spearman

correlations No evidence of Linder type effect

Kennedy and

McHugh (1983) 1963, 1970,

1976 United States Regression

analysis No support for Linder hypothesis nor H-O

theory Arnon and

Weinblatt

(1998) 1991 35 countries Bilateral gravity

trade model In favour of the Linder hypothesis for both

developed and less developed countries

McPherson,

Redfearn and

Tieslau (2001) 1984-1992

6 developing

East African

countries

Cross-section

time-series panel

data

An evidence of the Linder effect was

found for 5 countries (Ethiopia, Kenya,

Rwanda, Sudan and Uganda)

Choi (2002) 1970, 1980,

1990, 1992 63 countries Gravity type

regression model A favourable result was obtained in

support of the Linder hypothesis

Bukhari et al.

(2005) 1993-2002 Three South

Asian countries

Cross-section

time-series panel

data

Strong evidence in favour of the Linder

hypothesis for aforementioned three

countries

Bohman and

Nilsson (2007) 2000 57 countries Income overlap The results support the Linder hypothesis

with strongest results for differentiated

products

Haq and Meilke

(2008) 1990-2000

52 developed

and

developing

countries

OLS and Tobit

estimation No evidence of Linder effect in

differentiated agri-food product trade

Hallak (2010) 1995 64 countries Sectoral OLS and

ML estimates Support for the sectoral Linder hypothesis

Rauh (2010) 2002-2007 Germany Panel data,

country and time

fixed effects

Results reaffirm the Linder hypothesis for

Germany in trade with other European

countries

Jian (2011) 2000-2009 China and EU

countries

Gravity model

and panel data

analysis

Sign of the Linder hypothesis - countries

trade more if the gap of per capita GDP

between them is smaller

Bo (2013) 2001-2010 China and its

fourteen

trading partners

Gravity model

approach Linder hypothesis is supported

Kahram (2014) 1992-2012 Iran and its

bilateral trading

partners

Fixed-Random

effects model

On the one side, for some groups the

evidence of the Linder effect was not

found. On the other side, in the low-

income country group, the effect is very

strong and elastic

Atabay (2015) 1996-2010 BRIC countries Gravity equation

panel data

Trade between BRIC countries is getting

bigger over the years and the evidence

of the Linder hypothesis between these

countries has been found

Steinbach

(2015) 1995-2012

152 countries,

737 agricultural

and food

products

Sectoral gravity

equation

Similar aggregate preferences are

important determinant of bilateral export

trade in agricultural and food products -

evidence in support of Linder hypothesis

Niem (2016) 1994-2004

China's

cosmetic

industry and its

14

representative

trading partners

OLS fixed effects Empirical tests support the Linder

hypothesis

Source: Authors’ collection.

Bohman and Nilsson (2007) invented methodology which takes into account

income overlap between countries when testing the Linder hypothesis. Instead of

using difference between gross domestic product per capita as Linder variable, their

model includes distribution of income in countries.

They identify common market between countries by calculating income overlap

while Linder variable is formed in two ways; common market in relation to home

market and absolute size of the market. They found positive effect of the Linder

variables on export intensity supporting the Linder hypothesis.

66



Vol. 4, No. 1, 2018, pp. 62-73

Haq and Meilke (2008) tested Linder effect in differentiated agri-food product

trade using generalized gravity equation. Linder variable was approximated as

absolute difference between trading countries gross domestic product per capita

and Balassa index. Data were available for the period between 1990 and 2000 on

the sample of 52 countries. Hallak (2010) analysed Linder prediction of trade on a

sectoral level. He differentiated between high quality goods and low quality goods

and three sectoral categories. He failed to provide support for Linder effect using

gravity model on data for 64 countries in the year1995. When analysis was

conducted on a sectoral level the results were in favour of Linder hypothesis. Rauh

(2010) inspected trade pattern between Germany and European Union countries in

the period from 2002 to 2007. He used panel data regression model with time and

country-fixed effects. Results affirmed the Linder hypothesis indicating Germany’s

orientation on exports to EU countries with similar demand structures. Jian (2011)

investigated the Linder hypothesis for China and EU and came to the conclusion

that there is a sign of Linder effect. The smaller the gap of GDP per capita between

China and EU member country, the bigger trade volume becomes. In his model, he

also incorporated border effect, endowment and economic size.

Bo (2013) checked Linder hypothesis in the case of China's bilateral trade with its

trading partners on the trade data in the period 2001-2010. The fourteen countries

that are used in this paper are countries with the highest trading value for China.

Gravity model was applied in a panel regression analysis using explanatory variables

gross domestic product, differential gross domestic product per capita, real

exchange rate, population and distance. The conclusion of the analysis is that Linder

hypothesis for China's bilateral trading volume is supported while fixed effects model

performed better than random effects model. Kahram (2014) came to conclusion

that the Linder effect is very strong for particular groups of study for bilateral trade of

Iran, but is not the only factor that affected direction and trading volume of Iran. It is

also strongly affected by political factors such as ideology of the Islamic Republic of

Iran, international sanctions and economic factors such as differences in economic

development between countries, geographical distance between them, common

language and borders, etc.

Atabay (2015) found an evidence of Linder hypothesis for BRIC (Brazil, Russia, India

and China) countries on data from 1996 to 2010. Trade between these countries is

getting bigger over the years and, according to author’s belief, BRIC countries have

everything they need to become the biggest economies in the world. The novelty in

this study is implementation of crisis dummy variable. Steinbach (2015), using a

sample of 152 countries, tested the Linder hypothesis for bilateral trade in agricultural

and food products. The paper, in which 737 agricultural and food products were

taken in consideration, provided the little empirical evidence in support of the theory

and showed that similar aggregate preferences are important determinant of

bilateral export trade in agricultural and food products. Niem (2016) linked concepts

of intra-industry trade and Linder hypothesis. He measured intra-industry trade levels

using Grubel-Lloyd index. On the data for China's cosmetic industry on quality

differenciated products and its 14 representative trading partners in the period from

1994 to 2004 he came to the conclusion that there is an evidence of the Linder

effect.

Methodology and data In order to investigate trade pattern of Croatia's international trade, panel regression

model is constructed including 184 Croatia's import partner countries in the period

from 2000 to 2016. Panel regression analysis is used in order to take advantage of

67



Vol. 4, No. 1, 2018, pp. 62-73

cross-section and time-series property of data. All available data for imports are used

including those where are no trade from exporting country to Croatia. Omitting

these data would lead to biased results and therefore false conclusions about

validity of Linder hypothesis in the case of Croatia. Imports are observed as total

imports and imports of manufactured products. Manufactured imports are better

suited to estimate the validity of Linder hypothesis as stated in original Linder paper.

Dependant variable in regression is imports in Croatia while explanatory variables

are Linder variable, gross domestic product, distance and dummy variables

representing common border, EU and CEFTA membership. Alongside Linder variable,

it can be noticed enrollment of the variables GDP, distance and common border

describing gravity trade model. Gravity model is introduced by Dutch economist Jan

Tinbergen in 1962 (Tinbergen, 1962). It explains the international trade similar to

Newton's law of gravity. Trade is proportional to economic sizes of countries and

inversely proportional to distance between them. Model is in linear form because

log-log or log-linear model could not be made due to numerous zeros for non-trade

between trading partners. In equation 1 is presented cross-country panel regression

model: 𝐼𝑚𝑝𝑜𝑟𝑡𝑠𝑖𝑗𝑡 = 𝛽0 + 𝛽1𝐿𝑖𝑛𝑑𝑒𝑟𝑖𝑗𝑡 + 𝛽2𝐺𝐷𝑃𝑗𝑡 + 𝛽3𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗𝑡

+ 𝛽4𝐶𝑜𝑚𝑚𝑜𝑛_𝐵𝑜𝑟𝑑𝑒𝑟_𝐷𝑢𝑚𝑚𝑦𝑖𝑗𝑡 + 𝛽5𝐸𝑈_𝐷𝑢𝑚𝑚𝑦𝑖𝑗𝑡

+ 𝛽6𝐶𝐸𝐹𝑇𝐴_𝐷𝑢𝑚𝑚𝑦𝑖𝑗𝑡 + 𝜀𝑖𝑗𝑡 (1)

where 𝐼𝑚𝑝𝑜𝑟𝑡𝑠𝑖𝑗𝑡- variable denoting imports to Croatia 𝑖 from country 𝑗 in the period

from 2000 to 2016. Data for Croatia's total and merchandise trade are collected

from the World Bank (2018b); 𝐿𝑖𝑛𝑑𝑒𝑟𝑖𝑗𝑡- variable representing absolute difference

between exporting country 𝑗 GDP per capita and Croatia's GDP per capita. GDP

per capita values are expressed in PPP current international dollars. In order to

calculate Linder variable data for GDP per capita values for Croatia and exporting countries are provided from the World Bank (2018a); 𝐺𝐷𝑃𝑗𝑡- Gross Domestic Product

country 𝑗 (in current US$). Data for GDP values can be found at World Bank (2018);

𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗𝑡 - variable representing distance from Croatia's capital city Zagreb and

exporting country 𝑗 capital city. Data for distance variable are provided from the

DistanceFromTo (2018) and expressed in kilometres measured by air distance between capital cities; 𝐶𝑜𝑚𝑚𝑜𝑛_𝐵𝑜𝑟𝑑𝑒𝑟_𝐷𝑢𝑚𝑚𝑦𝑖𝑗𝑡 - dummy variable getting value 1 if

Croatia has common border with country j . Croatia has common border with 5

countries: Bosnia and Herzegovina, Hungary, Montenegro, Serbia and Slovenia; 𝐸𝑈_𝐷𝑢𝑚𝑚𝑦𝑖𝑗𝑡 - dummy variable getting value 1 if Croatia and exporting country are

both EU member countries. There are currently 28 European Union member countries

as of May, 1 2018. Croatia became 28th European Union member country on July

2013. Facts on EU enlargement process are available from European Commission

(2018); 𝐶𝐸𝐹𝑇𝐴_𝐷𝑢𝑚𝑚𝑦𝑖𝑗𝑡 - dummy variable getting value 1 if Croatia and exporting

country are CEFTA member countries. Croatia was CEFTA member country from 2003

to 2013, CEFTA Secretariat (2018).

Expected signs of regression are negative for Linder variable, distance and

positive for gross domestic product, common border dummy, EU and CEFTA

dummies. It is expected that higher value of gross domestic product is positively

correlated with higher value of imports in Croatia. On the other side, smaller

differences between gross domestic products per capita between Croatia and its

partner countries (Linder variable) should be associated with higher value of imports

in Croatia, according to Linder hypothesis. Larger distance between countries should

be correlated with lower value of mutual trade between partner countries. Non-zero

values of dummy variables (common border for Croatia's neighbour countries and

68



Vol. 4, No. 1, 2018, pp. 62-73

membership in regional economic integrations) should be associated with larger

value of imports in Croatia. Cross-country panel regression model is conducted using

pooled OLS (POLS), fixed effects (FE) and random effects (RE) models. Hausman test

is used when making a decision about suitable model respectively to choose

between fixed effects and random effects model. In order to rule out the near

singular matrix (or dummy variable trap) in the estimation, dummy variables

common_border dummy, EU and CEFTA dummy are excluded from the equation

regression for fixed effects model. Those variables do not vary within cross-sections

and are therefore not identified in a fixed effects specification. According to

(Kahram, 2014) variable distance should also not be used in fixed effects model as it

is time invariant. In addition, mixed fixed and random effects are not allowed for

unbalanced data so they are not specified in the analysis. Main results and

discussion of the paper are given below.

Results and discussion In Table 2 are presented descriptive statistics of variables total imports,

manufactured imports, Linder variable, gross domestic product (GDP), distance

variable, common border dummy, EU and CEFTA dummies included in the panel

regression analysis.

Table 2 Descriptive statistics of variables

Variable Total

imports

Manuf.

imports

Linder

variable GDP Distance

Common

border EU CEFTA

Mean 1.04E+08 72733253 13883.11 3.13E+11 5974.9 0.0329 0.0349 0.0158

Median 647851 81539 11837.67 1.87E+10 5316.4 0 0 0

Maximum 5.26E+09 4.03E+09 118356.0 1.86E+13 18259.9 1 1 1

Minimum 0 0 22.33732 13196545 117.4 0 0 0

Std. Dev. 3.89E+08 2.97E+08 11399.72 1.29E+12 4114.04 0.1785 0.1835 0.1248

Skewness 6.473048 7.170135 2.937078 8.9853 0.6285 5.2313 5.0679 7.7562

Kurtosis 52.91652 63.83129 16.79538 98.8287 2.9688 28.367 26.684 61.1589

Jarque-Bera 342823 503560 28982.8 1225495 203.8507 97071.14 85560 467078

Probability 0 0 0 0 0 0 0 0

Sum 3.23E+11 2.25E+11 42954349 9.69E+14 18486355 102 108 49

Sum Sq. Dev. 4.67E+20 2.72E+20 4.02E+11 5.17E+27 5.24E+10 98.6373 104.23 48.2239

Obs. 3094 3094 3094 3094 3094 3094 3094 3094

Source: Authors calculations

In Table 3 is displayed cross-country panel regression model for total trade

between Croatia and its partner countries. Panel is unbalanced with 3094

observations included. Dependant variable is total import (import_all) while

independent variables are Linder variable, gross domestic product, distance

variable, common border, EU and CEFTA dummies.

Panel regression model is estimated for pooled OLS, fixed and random effects

model. Fixed effects model could not be specified for all independent variables,

only for GDP and Linder variable, as reasoned in the methodology section.

Statistically significant independent variables in the regression under 1 percent level

of significance are GDP, distance, common border dummy, CEFTA and EU dummies

for both POLS and RE models. In the fixed effects model variable GDP is again

significant. All coefficients except coefficient for Linder variable are consistent with

expected signs of regression. Linder variable is not significant in the regression

indicating that Croatia did not prefer trade with countries of similar preference

structures and equal level of economic development. Adjusted R-squared for POLS is

0.34 meaning that the model is moderately well explained which is in the line with

previous researches. In Figure 1 is presented scatter diagram which represents

69



Vol. 4, No. 1, 2018, pp. 62-73

relationship between total imports and GDP variable. It can be noticed that the

relationship is positive; higher values of gross domestic products are correlated with

higher imports values.

Table 3 Cross-country panel regression for all goods Dependent variable IMPORTS_ALL Panel regression model

Independent variables POLS FE RE

Constant 1.33E+08

(9.859779)

7.04E+07

(7.911914)

1.38E+08

(3.599368)

LINDER -415.6058

(-0.825128)

941.9113

(1.553474)

100.6547

(0.175940)

GDP 8.90E-05***

(20.17322)

6.66E-05***

(9.438025)

6.83E-05***

(10.71319)

Distance -14568.31***

(-9.891849)

-15091.71***

(-2.992470)

Common_Border_Dummy 8.61E+08***

(25.45030)

8.25E+08***

(7.053763)

EU_Dummy 3.53E+08***

(11.06789)

1.64E+08***

(10.80979)

CEFTA_Dummy -2.48E+08***

(-5.230207)

55799673***

(2.356260)

Adjusted R-squared 0.341739 0.878680 0.093649

S.E. of regression 3.15E+08 1.40E+08 1.38E+09

Prob. (F-statistic) 0.00000 0.00000 0.00000

Mean dependent variable 1.04E+08 1.04E+08 12733182

S.D. dependent variable 3.89E+08 3.89E+08 1.45E+09

Akaike info criterion 41.97830 40.40478

Durbin -Watson 0.091375 0.442597 0.424298

Observations 3094 3094 3094

Correlated random effects (Hausman test) Chi-Sq. Statistic (2.410675) , Prob. (0.2996)

Source: Authors' calculations. OLS show White heteroskedasticity, standard errors and covariances are consistent, t-statistics is presented in

parentheses, *** significant at 1 percent, **significant at 5 percent and * significant at 10 percent level.

Figure 1 Total imports vs GDP variable

Source: Authors' calculations.

In Figure 2 is presented scatter diagram for relationship between total imports and

distance variable. Relationship is as expected negative indicating lower value of

imports as distance between trading countries is larger.

70



Vol. 4, No. 1, 2018, pp. 62-73

Figure 2: Total imports vs distance variable

Source: Authors' calculations

In Table 4 is presented cross-country panel regression for trade in manufactured

goods between Croatia and exporting countries. Linder hypothesis is originally

formulated for manufactured goods so this modification in the regression model

contributes in the positive way in estimation of the validity of the Linder hypothesis.

Table 4 Cross-country panel regression for manufactured goods Dependent variable IMPORTS_MAN Panel regression model

Independent variables POLS FE RE

Constant 76938892

(7.351848)

49280134

(8.100681)

94694492

(3.157316)

LINDER 432.8292

(1.103652)

699.3433

(1.687864)

297.8475

(0.747753)

GDP 7.32E-05***

(21.28987)

4.39E-05***

(9.100660)

4.67E-05***

(10.42843)

Distance -9837.867***

(-8.579168)

-10603.87***

(-2.663914)

Common_Border_Dummy 6.00E+08***

(22.77382)

5.71E+08***

(6.188568)

EU_Dummy 2.57E+08***

(10.36300)

87195816***

(8.332517)

CEFTA_Dummy -1.96E+08***

(-5.310161)

27353295*

(1.675840)

Adjusted R-squared 0.315351 0.896621 0.073256

S.E. of regression 2.46E+08 95400404 94884317

Prob. (F-statistic) 0.0000 0.0000 0.0000

Mean dependent variable 72733253 72733253 77396471

S.D. dependent variable 2.97E+08 98559915 98559916

Akaike info criterion 41.47783 39.64330

Durbin -Watson 0.070433 0.437722 0.406988

Observations 3094 3094 3094

Correlated random effects (Hausman test) Chi-Sq. Statistic (4.474261) , Prob. (0.1068)

Source: Authors' calculations. OLS show White heteroskedasticity, standard errors and covariances are consistent, t-statistics is presented in

parentheses, *** significant at 1 percent, **significant at 5 percent and * significant at 10 percent level.

71



Vol. 4, No. 1, 2018, pp. 62-73

Results are similar with previous ones for total goods trade. The exception is CEFTA

dummy variable which is significant under 10 percent of significance in RE model.

Chi-Square statistics for Hausman test is 4.474261 with probability of 0.1068 indicating

that random effects model is preferable over fixed effects model. According to null

hypothesis preferred model is random effects model while alternative hypothesis

chooses fixed effects model as suitable. In Figure 3 is presented scatter diagram for

relationship between manufactured imports and Linder variable. It can be seen that

imports did not depend on absolute difference between partner trading countries

gross domestic product per capita.

Figure 3: Manufactured imports vs Linder variable

Source: Authors' calculations.

It can be concluded that validity of Linder hypothesis cannot be accepted in the

case of Croatia, instead pattern of trade is in the line with gravity model of

international trade. Croatia imports relatively more from countries of larger

economic size than countries of similar level of gross domestic product. Distance is

also limiting factor that negatively affects volume of international trade.

Limitation of the analysis can be attributed to the use of absolute difference

measure of GDP per capita between trading countries as approximation for Linder

effect. It diminishes in a certain way the impact of larger economic size on importing

country. In addition to existing measures of Linder effect, such as overlapping

demands and Balassa index, we suggest the usage of relative difference between

gross domestic products between trading countries as approximation for Linder

variable. It should have better performance than absolute difference but further

investigations should be made taking into account different countries and different

time periods.

Conclusion Goal of this paper was to empirically investigate validity of Linder hypothesis in the

case of Croatia. Linder prediction is that the most of trade should occur between

countries of similar demand structures and similar level of economic development.

72



Vol. 4, No. 1, 2018, pp. 62-73

Dependant variable in regression was imports in Croatia while explanatory variables

were Linder variable, gross domestic product, distance and dummy variables

representing common border, EU and CEFTA membership. Panel regression model

was estimated using pooled OLS, fixed and random effects model. According to

Hausman test, random effects model was better suited to data. Linder variable was

not significant in the regression indicating that Croatia did not prefer trade with

countries of similar level of economic development and similar preference structures.

Therefore, the validity of Linder hypothesis cannot be accepted in the case of

Croatia. Structure of Croatian imports trade is more in the line with gravity model of

international trade; higher value of partner country's GDP per capita increases

mutual trade while distance between them diminishes it. Contribution of research is

comprising explanation of pattern of trade between Croatia and its trading partners

and adding to the existing literature about Linder hypothesis testings in developed

and developing countries. This can be useful for country policy makers when making

strategic long-term decisions about country trade policy. Limitation of the research is

related to the use of absolute difference measure of GDP per capita between

trading countries as approximation for Linder effect which diminishes the importance

of economic size.

References 1. Arnon, A., Weinblatt, J. (1998). Linder's hypothesis revisited: income similarity effects for low

income countries. Applied Economics Letters, Vol. 5, No. 10, pp. 607-611.

2. Atabay, R. (2015). Validity of Linder Hypothesis in BRIC countries. Journal of International

Trade, Logistics and Law, Vol. 1, No. 2, pp. 85-92.

3. Bo, P. (2013). Bilateral trade of China and the Linder hypothesis: A gravity model

approach. Bachelor Thesis within Economics. Available at https://www.diva-

portal.org/smash/get/diva2:605666/FULLTEXT01.pdf [01 May 2018].

4. Bohman, H., Nilsson, D. (2007). Market overlap and the direction of exports: A new

approach of assessing the Linder hypothesis. Available at:

https://ideas.repec.org/p/hhs/cesisp/0086.html [01 May 2018].

5. Bukhari, S., Ahmad, M., Alam, S., Bukhari, S., Butt, M. (2005). An Empirical Analysis of the

Linder Theory of International Trade for South Asian Countries. The Pakistan Development

Review, Vol. 44, No. 3, pp. 307-320.

6. CEFTA Secretariat (2018). CEFTA Secretariat. Available at http://cefta.int/ [01 May 2018].

7. Choi, C. (2002). Linder Hypothesis Revisited. Applied Economics Letters, Vol. 9, No. 9pp.

601-605.

8. DistanceFromTo (2018). Distance Between Cities Places On Map. Available at

https://www.distancefromto.net/ [01 May 2018].

9. European Commission (2018). EU enlargement factsheet: How does the EU accession

process work? Available at https://ec.europa.eu/neighbourhood-

enlargement/sites/near/files/pdf/publication/factsheet_en.pdf [01 May 2018].

10. Hallak, J. C. (2010). A product-quality view of the Linder hypothesis. The Review of

Economics and Statistics, Vol. 92, No. 3, pp. 453-466.

11. Haq, Z., Meilke, K. D. (2008). Differentiated Agri-Food Product Trade and the Linder Effect.

Canad. Agricultural Trade Policy Research Network Working Paper, No. 2008-07, pp. 1-26.

12. Heckscher, E. F. (1919). The effect of foreign trade on the distribution of income.

Ekonomisk Tidskrift, Vol. 21, pp. 497-512. Reprinted in Ellis, H. S., Metzler, L. A. (eds.).

Readings in the Theory of International Trade. Allen and Undwin, London.

13. Jian, Z. (2011). Based on gravity trade model and Linder Hypothesis: An Empirical

Application to China-EU Trade Flows. Proceedings of 2011 International Conference on

Business Management and Electronic Information, 13-15 May 2011, Guangzhou, China.

14. Kahram, A. (2014). The Comparative Analysis of the Linder Hypothesis: The Bilateral Trade

Model between Iran and Its Trade Partners. Advances in Economics and Business

Management (AEBM), Vol. 1, No.1, pp. 1-8.

73



Vol. 4, No. 1, 2018, pp. 62-73

15. Kennedy, T., McHugh R. (1980). An Intertemporal Test and Rejection of the Linder

Hypothesis. Southern Economic Journal, Vol. 46, No. 3, pp. 898-903.

16. Kennedy, T., McHugh, R. (1983). Taste Similarity and Trade Intensity: A Test of the Linder

Hypothesis for U. S. Exports. Weltwirtschaftliches-Archiv, Vol. 119, No. 1, pp. 84-96.

17. Leontief, W. (1953). Domestic Production and Foreign Trade: The American Capital

Position Re-Examined. Proceedings of the American Philosophical Society, Vol. 97, pp.

332-349. Reprinted in Richard, C., Harry, G. J. (eds.). Readings in International Economics.

Homewood.

18. Linder, S. B. (1961). An Essay on Trade and Transformation. Wiley and Sons, New York.

19. McPherson, M. A., Redfearn, M. R., Tieslau M. A. (2001). International Trade and

Developing Countries: An Empirical Investigation of the Linder Hypothesis. Applied

Economics, Vol. 33, No. 5, pp. 649-657.

20. Niem, L. D. (2016). Linder Hypothesis and Trade of Quality Differentiated Good: A Case of

Cosmetic Industry of China. Modern Economy, Vol. 7, No. 3, pp. 307-313.

21. Ohlin, B. (1933). Interregional and International Trade. Harvard University Press,

Cambridge.

22. Rauh, A. (2010). Empirical analysis of the Linder hypothesis: The case of Germany's trade

within Europe. The American Economist, Vol. 55, No. 2, pp. 136-141.

23. Ricardo, D. (1817). The principles of political economy and taxation. John Murray, London.

24. Smith, A. (1776). An inquiry into the nature and causes of the wealth of nations. Alex

Catalogue, Raleigh.

25. Steinbach, S. (2015). Bilateral export trade and income similarity: Does the Linder

hypothesis hold for agricultural and food trade? AgEcon Search, Research in Agricultural

& Applied Economics, Conference Paper 2015-05, pp. 110-114.

26. Tinbergen, J. (1962). Shaping the World Economy. Suggestion for an International

Economic Policy. New York.

27. Trefler, D. (1995). The Case of the Missing Trade and Other Mysteries. American Economic

Review, Vol. 85, No. 5, pp. 1029-1046.

28. World Bank (2018). GDP (Current US$). Available at

https://data.worldbank.org/indicator/NY.GDP.MKTP.CD [01 May 2018].

29. World Bank (2018a). GDP per capita (Current US$). Available at

https://data.worldbank.org/indicator/NY.GDP.PCAP.CD [01 May 2018].

30. World Bank (2018b). World Integrated Trade Solution. Available at

https://wits.worldbank.org/CountryProfile/en/Country/HRV/StartYear/2000/EndYear/2016/

TradeFlow/Import/Partner/ALL/Indicator/MPRT-TRD-VL [01 May 2018].

About the authors Hrvoje Jošić has been employed at the Faculty of Economics and Business, University of

Zagreb since 2006. In 2011 he gained a doctoral degree at the Faculty of Economics and

Business, University of Zagreb. He currently holds the title of an Assistant Professor and works at

the International Economics Department at the Faculty of Economics and Business in Zagreb.

He teaches the International Economics course. His main research interests are theories of

international trade, the balance of payment and the debt problem. He is an author of one

book and many scientific papers in international journals and conference proceedings. He

can be contacted at: [email protected].

Matej Metelko is a 3rd year student at the Faculty of Economics and Business in Zagreb. He is

a student assistant at the International Economics Department, which is one of his major

interests, and the finance team leader in the student association Economic Clinic. He has

also attended the Sport Management course at Bocconi Summer School. He can be

contacted at: [email protected].

http://www.scirp.org/journal/ME

mailto:[email protected]

testing the validity of the linder hypothesis for croatia...leontief (1953) found paradox in the...

Documents